CODEWITHSUNDEEP
rmatrixreshape
jwillis0720
jwillis0720

Reputation: 4477

Reconstruct symmetric matrix from values in long-form

I have a tsv that looks like this (long-form):

  one   two   value
  a     b     30
  a     c     40
  a     d     20
  b     c     10
  b     d     05
  c     d     30

I'm trying to get this into a dataframe for R (or pandas)

    a  b  c  d 
a   00 30 40 20
b   30 00 10 05 
c   40 10 00 30
d   20 05 30 00

The problem is, in my tsv I only have a, b defined and not b,a. So I get a lot of NAs in my dataframe.

The final goal is to get a distance matrix to use in clustering. Any help would be appreciated.

Upvotes: 10

Views: 3222

Answers (6)

ThomasIsCoding
ThomasIsCoding

Reputation: 101335

With base R, you can try xtabs like below

xtabs(
    value ~ .,
    rbind(
        df,
        setNames(df, names(df)[c(2, 1, 3)])
    )
)

which gives

   two
one  a  b  c  d
  a  0 30 40 20
  b 30  0 10  5
  c 40 10  0 30
  d 20  5 30  0

Upvotes: 0

Jeff Bezos
Jeff Bezos

Reputation: 2253

The trusty for loop can sometimes be the most intuitive:

# Simulate data.
df <- data.frame(one = c("a", "a", "a", "b", "b", "c"), 
                 two = c("b", "c", "d", "c", "d", "d"), 
                 value = c(30, 40, 20, 10, 5, 30))

# Initialize matrix.
cols <- unique(c(df_long$one, df_long$two))
mat <- matrix(nrow = length(cols), ncol = length(cols),
              dimnames = list(cols, cols))

# Populate matrix.
for(i in seq(nrow(df))) {
    mat[df$one[i], df$two[i]] <- df$value[i]
    mat[df$two[i], df$one[i]] <- df$value[i]
}
diag(mat) <- 0

Upvotes: 0

smci
smci

Reputation: 33938

Yet another approach is reshape::cast

df.long = data.frame(one=c('a','a','a','b','b','c'),
                     two=c('b','c','d','c','d','d'),
                     value=c(30,40,20,10,05,30) )

# cast will recover the upper/lower-triangles...
df <- as.matrix( cast(df.long, one ~ two, fill=0) )
#    b  c  d
# a 30 40 20
# b  0 10  5
# c  0  0 30

So we construct matrix with full indices, and insert:

df <- matrix(nrow=length(indices), ncol=length(indices),dimnames = list(indices,indices))    
diag(df) <- 0
# once we assure that the full upper-triangle is present and in sorted order (as Robert's answer does), then we
df[upper.tri(df)] <- as.matrix( cast(df.long, one ~ two, fill=0) )
df[lower.tri(df)] <- df[upper.tri(df)]

UPDATE: the original sketch included these manual kludges

Then the same approaches to add the missing row 'd' and column 'a', and fill the lower triangle by adding the transpose t(df) :

df <- cbind(a=rep(0,4), rbind(df, d=rep(0,3)))
#   a  b  c  d
# a 0 30 40 20
# b 0  0 10  5
# c 0  0  0 30
# d 0  0  0  0

df + t(df)
#    a  b  c  d
# a  0 30 40 20
# b 30  0 10  5
# c 40 10  0 30
# d 20  5 30  0

Upvotes: 2

user20650
user20650

Reputation: 25854

An igraph solution where you read in the dataframe, with the value assumed as edge weights. You can then convert this to an adjacency matrix

dat <- read.table(header=T, text=" one   two   value
  a     b     30
  a     c     40
  a     d     20
  b     c     10
  b     d     05
  c     d     30")

library(igraph)

# Make undirected so that graph matrix will be symmetric
g <- graph.data.frame(dat, directed=FALSE)

# add value as a weight attribute
get.adjacency(g, attr="value", sparse=FALSE)
#   a  b  c  d
#a  0 30 40 20
#b 30  0 10  5
#c 40 10  0 30
#d 20  5 30  0

Upvotes: 9

akrun
akrun

Reputation: 887108

You may try

 un1 <- unique(unlist(df1[1:2]))
 df1[1:2] <- lapply(df1[1:2], factor, levels=un1)
 m1 <- xtabs(value~one+two, df1)
 m1+t(m1)
 #    two
 #one  a  b  c  d
 #a    0 30 40 20
 #b   30  0 10  5
 #c   40 10  0 30
 #d   20  5 30  0

Or you use the row/col index

  m1 <- matrix(0, nrow=length(un1), ncol=length(un1),
                              dimnames=list(un1, un1))
  m1[cbind(match(df1$one, rownames(m1)), 
               match(df1$two, colnames(m1)))] <- df1$value
  m1+t(m1)
  #   a  b  c  d
  #a  0 30 40 20
  #b 30  0 10  5
  #c 40 10  0 30
  #d 20  5 30  0

Upvotes: 2

Robert
Robert

Reputation: 5152

Make sure your data is sorted tsv=tsv[with(tsv,order(one,two)),], and try this:

n=4
B <- matrix(rep(0,n*n), n)
dimnames(B) <- list(letters[1:n],letters[1:n])
B[lower.tri(B)] <- tsv$value
B[upper.tri(B)]=tsv$value
B

Upvotes: 1

Related Questions